Geochemistry of Lavas, Pumice and Veins in Drill Core GPK-1, Palaea Kameni, Santorini

The section 0-155 m consists of porphyritic and vesicular dacitic lavas, underlain by pyroclastics (or pumiceous matter) interbedded with thin layers of latitic andesitic lava. The rocks show apparent alterations, mainly in the interval 7-54 m, where bleached and corroded sections and ferric oxide coated veins are common.

Major and trace element variations show that the lavas belong to three units, C-lavas at 0-125 m being the youngest, B-lavas at 125-153 m, and the oldest A-lavas at 153-202 m; this layer also includes the pumiceous beds.

Classification diagrams show that the B- and C-lavas are dacites and rhyodacites, whereas the A-lavas are quartz latiandesites to quartz andesites, as are also the pyroclastic components. Very little alteration has taken place, e.g. some albitization, which is corroborated by data for the hydrothermal waters.

Primordial mantle-normalized fractionation plots and multicationic diagrams suggest that all lavas and pyroclastics have a common magmatic source differentiating to andesitic, dacitic and rhyodacitic rocks. The pumiceous sequence is admixed with much hydrothermal Fe-rich matter, which is strikingly similar to the red oxyhydroxide vein coatings in the upper lava section.

The lavas and pyroclastics, including the pumice of the Minoan eruption, are all of post-caldera age. The iron-enriched pumice beds show that hydrothermal processes have probably operated since the caldera formed.

INTRODUCTION

The geochemistry and provenance relations of the island arc related volcanic rocks and hydrothermal systems at Santorini (Fig. 1) have been studied by several authors (Butuzova 1969; Puchelt 1973; Smith and Cronan 1983; Boström and Widenfalk 1984). The genetic conditions at depth of such hydrothermal systems have been little studied, however.

In order to understand these problems better a drill core was procured under a joint Greek-Swedish drilling project (Arvanitides et al. 1988). We will here present the initial geochemical studies of the entire rock-section.

GEOLOGICAL SETTING OF THE HYDROTHERMAL SYSTEM

The Santorini islands resulted from volcanic activity of Late Pleistocene to recent times at the site where the African plate dips under the Aegean plate. The present caldera resulted from an eruption at about 1400-1600 BC, the volcanic Kameni islands surfacing much later (Hammer et al. 1980; Pichler and Kussmaul 1980; Heiken and McCoy 1984).

The Kameni islands consist of dacitic flows and domes rising a few hundred metres above the deepest part of the caldera floor (Huijsmans 1985); on these islands high-temperature (~ 35-40° C) springs and iron-silica-rich exhalative sediments occur in several bays.

FIELD WORK AND MAROSCOPIC CORE DESCRIPTION

In January 1988 the drilling on Palaea Kameni (Fig. 2) was terminated after having reached a depth of 201.5 m (Arvanitides et al. 1988) in order to analyse the samples obtained and to evaluate the scientific justification for further drillings.

At IGME the drill core was described and sampled at close intervals (Fig. 3). The rocks at 0-155 m show various massive, porphyritic and vesicular features and resemble the dacite-rhyodacite characteristic of the Kameni islands (Pichler and Kussmaul 1972; 1980; Huijsmans 1985). Macroscopically the rocks appear altered, particularly in the interval 7-54 m, as indicated by 1) abundant ferric oxide stained veins, 2) some bleached sections and 3) possibly some corrosion (?) of wall rocks. However, the recovery losses in this section may be due to comminution during the drilling of vesicular and glassy matter. Rubble zones are best developed where vesicular structures dominate the rocks.

Microscopic studies reveal that these dacitic rocks are porphyritic to glomeroporphyritic and mineralogically very uniform. The phenocrysts consist of plagioclase, clinopyroxene, orthopyroxene, and magnetite in a hyalopilitic groundmass. No microscopic evidence exists for any substantial hot water-rock interactions in the upper 150 metres of the drill hole (Paritsis et al. 1990).

Below the 155 m level the core samples primarily represent an unconsolidated, loose material, which is greyish, yellowish-grey or locally reddish-brown, due to ferric oxyhydroxide admixture. Visual inspection shows that much of the matter is pumiceous, present as a well-sorted sediment, whereas other fractions represent comminuted latite-andesite. At a few levels in this section there are thin lava layers, resembling latite-andesite in appearance. No post-depositional alterations were found in these rocks.

CHEMICAL ANALYSIS

During the core description 178 samples were collected for chemical analysis at SU (Stockholm University) and SGAB (Swedish Geological Co.). Due to iron oxide staining many samples had to be trimmed on a petrographic saw; after drying and grinding of the samples loss of ignition (LOI) was determined by heating to 1000° C. Subsequently major and trace element analyses were performed, using atomic emission spectroscopy with an inductively coupled argon plasma (ICP-AES) as excitation source. The analysed components include SiO₂, Al₂O₃, TiO₂, Fe₂O₃, MnO, MgO, CaO, Na₂O, K₂O, P₂O₅, Ag, Ba, Be, Co, Cr, Cu, La, Sc, Sr, Sn, V, Y, Yb, Zn, Zr, Ni, Mo, Nd, Th, W and Pb; lead data are semi-quantitative. Tests with standard rocks as unknown samples show that data for the major elements and for Ba, Cu, Sc, Sr, V, Y, Yb, Zn, Zr, Cr and Ni data are excellent; probably data for Nb are also of good quality. The number of determinations of Be is small, and the elements Mo, Th, W, Ag and Sn are either non-detectable or show much scatter, which may be real or due to poor reproducibility. Some on-going S, U and Th isotope studies will be reported elsewhere.

During the preparation of samples for analyses the ferric coatings proved to occur only as very thin stains, never exceeding 0.1 mm in thickness. Attempts to chip off material for analysis proved inefficient, leading to much rock contamination of the samples as revealed under the microscope. Therefore a petrographic saw was used to trim off 15 oxide coatings, including associated rock as thin slivers. The samples were then rinsed at 50° C for about 15-20 minutes in hydrochloric acid (1:3), to which was added hydroxylamin chloride as a reducing agent. This procedure removed the ferric oxyhydroxides efficiently, whereas remaining rock slivers showed no attack.

Due to the paucity of this hydroxide fraction a composite solution was made of all 15 leachings (see Table 1). The results are recalculated to an oxide and carbonate sum of 100%, since the weight losses of the leached specimens were not determined.

RESULTS

The results are summarized as mean values and standard errors (s.e.) in Table 1 and in Fig. 4 to 15. The s.e. values are convenient to use in a Students's t-test to check if data-sets are significantly different. The analyses of the fresh rocks indicate that all lavas vary little in composition. Our major element results show excellent agreement with data in Pichler and Kussmaul (1972, their Table 4), but some differences exist. Our TiO₂ data differ with some 30 (relative) % from the value by these authors. However, it should be kept in mind that present ICP-AES methods permit titanium analyses with relative standard deviations of about 0.7%; only a decade ago the corresponding error was nearly 20% (Burman and Boström 1979; Burman et al. 1978; Boström et al. 1990d). This conclusion is supported by the remarkable similarity between the fresh rocks in GPK-1, particularly then from the depth layer 153-201 m, and Nea Kameni rock data (Boström and Widenfalk 1984).

CHEMOSTRATIGRAPHIC SUBDIVISION

The high quality of the data for the major elements, and for Ti, Zr, Sc and Yb, permits a petrogenetic grouping of the results, using methods similar to those by Winchester and Floyd 1976 and by Pearce and Cann 1973. It is obvious that the lavas belong to three different units, one including the youngest layer 0-125 m (marked as C), another 125-153 m (B) and a third, oldest layer at 153-202 m (A) (see Fig. 4 and 5). The low values for Zr/TiO₂ and Yb/TiO₂ in the interval 0-20 m could indicate the existence of a fourth shallow flow unit, but this conclusion is statistically uncertain.

These boundary depths have been used for the calculations of the mean compositions given in Table 1; there and in the following, the seemingly altered rocks are indicated as 'altered'. The pyroclastics at 155-200 m are reported as a separate unit in the Table, as are also the vein coatings. The scatter in the pyroclastics data is much larger than for 'altered' rocks (Fig. 6, 7). Major and trace LIL elements and the REE (La, Yb and Y) show normal variations in the core, as do also the restricted ranges of K₂O/Na₂O (0.4-0.5), Ba/Sr (2-3) or the chondrite normalized values for La/Yb (3.0-3.8). The transition elements Cr, Ni, V, Co as well as Cu, Zn and Pb are enriched in the pumiceous matter (Table 1; Fig. 6, 7). Two-variable plots of siderofile elements (Fig. 8) show that pyroclastics are clearly separated from the B- and C-lavas, which have very restricted compositional ranges, in contrast to the pyroclastics. The A-lavas plot close to the previous groups (Fig. 8), on the extension of the trend for the B- and C-lavas. However, primordial mantle normalized Ba-K-Nb-La-Sr-P-Zr-Ti-Y-Yb fractionation plots (Fig. 11) of the pyroclastics and all lavas show concomitant patterns, indicating a common source of origin.

ROCK CLASSIFICATION

The distribution of the major elements SiO₂ versus K₂O (Fig. 10) indicates that the A- and C-lavas show predominantly dacitic compositions, the former closer to an andesitic character, whereas B-lavas are rhyolitic, and the pyroclastics span an andesite-dacite range. Plots of Zr/TiO₂ versus SiO₂ and of Zr/TiO₂ versus Nb/Y confirm these conclusions. A multi-cationic Q-P plot (Fig. 11) indicates that the B- and C-lavas have dacitic compositions, whereas the pyroclastics and the A-lavas have a quartz andesitic to quartz latiandesitic composition.

DISCUSSION

     The origin of the lavas and their alterations:     

Plots of normalized LIL elements and the rare earths (REE) Y, La and Yb fractionations (Fig. 9) suggest that all lavas and pyroclastics derive from a common magmatic source, and that differences in element concentrations are caused by the subsequent magmatic fractionation process and the imposed hydrothermal activity. More complete REE data are presently being compiled to further support the conclusions.

The C-lavas and perhaps also the B-lavas belong to the 197 BC (Fig. 10)post-caldera calc-alkaline volcanism at Santorini (Huijsmans 1985). The A-lavas are much older, since they alternate with pumiceous matter (Fig. 10) which represents erosion products from the caldera walls, which started to fill the caldera depression very soon after the last Minoan eruption and the concomitant caldera formation.

The A-, B- and C-lavas define a magmatic trend (Fig. 8), as indicated by the characteristic minerals biotite, hornblende, orthopyroxene, clinopyroxene, primary epidote and sphene, see multi-cationic A-B plot in Fig. 12. The lavas define an inverse compositional trend upward in the stratigraphy, with the dacitic-quartz latiandesitic A-lavas being replaced by the rhyodacitic-rhyolithic layer B, which is succeeded by the dacitic C-lavas. This trend indicates that the flows represent different eruptive cycles evolving from a common magma chamber rather than being differentiated from a single eruptive cycle (Huijsmans 1985; Kalogeropoulos and Paritsis 1990).

The mean compositions of the 'altered' rocks differ only little from the fresh rocks in the same unit (Table 1), corroborating the finding that the alteration is small (Paritsis et al. 1990). Some albitization tendencies are revealed by the geochemistry of the rocks (Fig. 13) and are clearly noticed in the hot-spring waters (Boström et al. 1990c). The 'altered' materials have larger LOI values and compositional scatter (standard errors) than the fresh rocks. This suggests that some alterations involve a local exchange of matter, leaving the bulk composition for a major rock unit little changed. An analysis, using Ti and Zr normalized data (Rona et al. 1980), indicates that the 'altered' rocks cannot be an important source of iron and other elements in the hydrothermal products at the Kameni islands.

     The origin of the pyroclastics:

The pyroclastic sediments at 155-201 m show much more compositional scatter even than the 'altered' rocks (Fig. 6); it is well known that extreme variations in sediments of, for instance, silica, carbonate and iron content are generally much larger than in igneous rocks. However, the overall composition of this layer is remarkably similar to that of pumice (Fig. 10), and the colour and the grain shape, density, etc., of much matter in this layer is identical to pumice collected on the main island of Thera.

A major feature of these sediments is the reddish-brown colour of some layers; chemical analyses show that these sections are enriched in iron (Fig. 15) compared to ordinary pumice, as is also the case for present caldera sediments (Butuzova 1969; Smith and Cronan 1978; Petersen and Müller 1978; Boström et al. 1990d). The iron content in the pumice follows the trend described for the hot-spring bays by Boström et al. (1984; 1990a) and the admixed iron phase shows distinct resemblances to the composition of the hot waters (Boström et al. 1990b, c) and the vein coatings observed in drill hole GPK-1, see Table 1 and Fig. 14.

Silica, alumina, iron and manganese abundances in the pumiceous section have varied with time, however (Fig. 6), iron contents showing maxima of about 12% Fe₂O₃ at the 165 and 175 metre levels, whereas the level 180-195 m largely shows values that scatter near 6%. This indicates that the rate of deposition of pumice and exhalative sedimentary products may have varied considerably with time. Lack of age determinations prevents a more detailed discussion, but it appears likely that the supply rates of silica may have varied much due to climatic effects on the erosion on the main island Thera.

It is obvious that the behaviour of the LIL elements was exclusively controlled by magmatic differentiation processes, whereas the pyroclastic layer was strongly enriched in iron compounds, precipitated from hydrothermal solutions. This led to an enrichment of compatible transition elements in the pumice, such as Cr, Ni and V, a trend that deviates from the magmatic trend for the lavas (Fig. 10). These sediments formed before Palaea Kameni surfaced, that is the hydrothermal processes have lasted two millennia, and probably since the formation of the caldera some 3500 years ago; hydrothermal systems of higher age are well known (Silberman et al. 1979; White et al. 1988).

     Models for the metallogenesis and the hydrothermal system:

Vein coatings and hot spring solutions in the most active hot-spring bay on Nea Kameni (at A in Fig. 1) show striking similarities. Iron, Al, P, Cr, Cu and Zn are present in similar relative abundances, but the vein coatings are poorer in Mn and Ba, and particularly in Ca (Fig. 14). However, this deficiency is predictable in view of the increased solubility of Mn, Ba and Ca in acid reducing systems. The vein coatings are therefore not descended weathering products; also, most vein matter should then be at the very top (0-7 m) of the drill core. The location of the vein deposits at the 7-54 m level hence suggest that this was the zone of optimal admixture to the solutions, but was not the main source region for the metalliferous solutions. It is more likely that these solutions derive from processes at great depths controlled by plate tectonic setting (Zelenov 1972; Boström et al. 1990b, c).

Waters at the bottom of the drill hole are considerably enriched in carbon dioxide, Si, Ba, Fe and Mn, and much warmer (24° C) than sea water (< 18° C) at that level (Boström et al. 1990c). These findings and the resemblance between a) the ferric oxyhydroxide veins, b) the hydrothermal sediments, c) the waters at the bottom of the drill hole and d) the hot-spring solutions at Nea and Palaea Kameni suggest a common deep source, located on a major tectonic feature between the Kameni islands and the submarine volcano Kolombos (Boström et al. 1990d).

Little is known about the chemical processes that generated the hydrothermal solutions. Harder (1964) pointed out that the carbon dioxide solutions could carry several metals, but numerous hot springs are known with abundant carbon dioxide but no metalliferous solutions (see review by Boström et al. 1990c). Boström and Widenfalk (1984) speculated that oxygen-containing waters at great depth mix and react with sulphide-rich volcanic emanations, forming sulphuric acid-rich solutions, that attack the local rocks. Since dacites contain only small amounts of trace-metals such as Mn, Zn, Ba and Cu, the solutions become poor in these constituents compared to Fe. An enigma is the behaviour of Al, but possibly this component redeposits at depth; this process has not yet been observed.

Present measurements of the thermal profile in GPK-1 show lower temperatures at depth, which is opposite to the pattern generally observed in thermal areas. This could be due to incomplete restoring of pre-drilling conditions, but another explanation would be that inflow of cold sea water at depth dominates over the flow of rising hot water (see site B in Fig. 16). Such flow patterns have been suggested for hydrothermal systems in the deep sea (Rona et al. 1983), and agree with known flow patterns in sea water (Sverdrup et al. 1942). The rising hot limb (RHL) may therefore be more narrow than the reactive zone (RZ) where sea water, hot magma and volcanic gases inter-react. The GPK-1 drill hole is therefore most probably located on the flank of the RHL.

The uniformity of the extreme types of the hydrothermal sediments and solutions, the temperatures in the waters and the geographical distribution of these phenomena all suggest that the depth of formation of the ultimate waters may be fairly deep down (Boström et al. 1990b, d). This drill site study yields the same conclusion, and demonstrates that future drillings must reach much further down than GPK-1 and be closer to an active centre.

CONCLUSIONS

1. The lavas of Palaea Kameni are all less than 3500 years old, since they rest on pumice beds of post-caldera formation age, the youngest lavas having formed at 197 BC. All lavas are rhyodacitic and dacitic and probably represent melts from a common magma source at different eruption cycles.
2. The ferric oxyhydroxide coatings in veins in the lavas represent deposits from a hydrothermal solution of the same type that presently debouches in hot springs on the Kameni islands.
3. Geochemical and mineralogical data indicate that all alterations are small in the uppermost 200 metres of the lavas, except for some corrosion of the vein walls; this corrosion is not a major source of matter in the hot-spring solutions, however.
4. The underlying pumice beds predate the lava flows and may represent early post-caldera formation products, formed when the young caldera walls were still highly unstable and quickly eroded, filling the caldera depression with excessive quantities of pumice and other pyroclastic matter. These pumice and pyroclastic beds originally had a primarily dacitic-latiandesitic composition, largely identical to the compositions of minor flow of lavas in this section.
5. Simultaneously with the formation of these pumice beds, hydrothermal deposition of iron-rich matter started; the formed products fall on the same geochemical evolution trends that have been found for other mixtures of pumice and exhalative phases in the Kameni islands and in the caldera of Santorini. This shows that hydrothermal processes started more than 2000 years ago and probably began right after the formation of the caldera about 1400-1600 BC. The presence of old hot springs in other caldera features, e.g. at Yellowstone (White et al. 1988) corroborates this high age for the hydrothermal processes at Santorini.
6. The drill hole studied here is probably located on the flank of a major rising hot water limb.

--------------------------------------------

--------------------------------------------